Genome-Wide Analysis of Alternative Splicing during ... · Genome-Wide Analysis of Alternative Splicing during Development and Drought Stress in Maize1[OPEN] Shawn R. Thatcher, Olga

Genome-Wide Analysis of AlternativeSplicing during Development andDrought Stress in Maize1[OPEN]

Shawn R. Thatcher, Olga N. Danilevskaya, Xin Meng, Mary Beatty, Gina Zastrow-Hayes, Charlotte Harris,Brandon Van Allen, Jeffrey Habben, and Bailin Li*

DuPont Pioneer, Wilmington, Delaware 19880 (S.R.T., B.L.); and DuPont Pioneer, Johnston, Iowa 50131 (O.N.D.,X.M., M.B., G.Z.-H., C.H., B.V.A., J.H.)

ORCID IDs: 0000-0002-3996-6617 (O.N.D.); 0000-0002-0092-6845 (B.V.A.).

Alternative splicing plays a crucial role in plant development as well as stress responses. Although alternative splicing has beenstudied during development and in response to stress, the interplay between these two factors remains an open question. Toassess the effects of drought stress on developmentally regulated splicing in maize (Zea mays), 94 RNA-seq libraries from ear,tassel, and leaf of the B73 public inbred line were constructed at four developmental stages under both well-watered anddrought conditions. This analysis was supplemented with a publicly available series of 53 libraries from developing seed,embryo, and endosperm. More than 48,000 novel isoforms, often with stage- or condition-specific expression, wereuncovered, suggesting that developmentally regulated alternative splicing occurs in thousands of genes. Drought inducedlarge developmental splicing changes in leaf and ear but relatively few in tassel. Most developmental stage-specific splicingchanges affected by drought were tissue dependent, whereas stage-independent changes frequently overlapped between leafand ear. A linear relationship was found between gene expression changes in splicing factors and alternative spicing of othergenes during development. Collectively, these results demonstrate that alternative splicing is strongly associated with tissuetype, developmental stage, and stress condition.

After transcription, the majority of eukaryotic pre-mRNA is subjected to a series of posttranscriptionalmodifications, including the removal of introns to forma mature mRNA (Stamm et al., 2005). Although someintrons are removed constitutively, many can beprocessed in a variety of alternativeways, including exonskipping, intron retention, alternative acceptor, alterna-tive donor, and alternative position (change in both ac-ceptor and donor positions; Lorkovi�c et al., 2000). Thesealternative splicing events form a crucial regulatory leveland have the ability to alter an mRNA’s stability, locali-zation, and protein products. Alternative splicing eventsare controlled by a variety of cis-elements, including thepresence of consensus splice sequences at the intron-exonborder and intronic and exonic splicing enhancer se-quences (Pertea et al., 2007). Trans-acting factors also exert

a strong effect on splicing and are mostly composed ofSer/Arg-rich proteins, which typically promote intronremoval, and heterogenous nuclear ribonucleoproteins,which typically inhibit it (Erkelenz et al., 2013). A host ofother indirect factors can affect splicing, including tran-scription rate, methylation status, and any cellular con-ditions that alter RNA secondary structure (Kornblihttet al., 2004; Shepard and Hertel, 2008; Regulski et al.,2013). These different factors combine to determinethe suite of isoforms that a gene produces in a giventissue, developmental stage, and environment.

Alternative splicing has been shown to play a role ina variety of different plant processes, including tissueidentity (Staiger and Brown, 2013). Genes involvedin processes ranging from auxin biosynthesis to irontransport have tissue-specific alternative splicing pat-terns that affect both protein localization and function(Kriechbaumer et al., 2012; Li et al., 2013). Tissue-specific splicing patterns can also alter a transcript’ssensitivity to regulation by microRNAs, such as themaize (Zea mays) SPX family, which produces miR827-sensitive isoforms that predominate in developingseedlings, while insensitive isoforms dominate othertissues (Thatcher et al., 2014). Gene expression changesin splicing factors also play key roles in guiding tissue-specific developmental processes, including seed de-velopment, where the U2AF splicing factor ROUGHENDOSPERM3 has been shown to be involved inmediating the interaction between the embryo and en-dosperm (Fouquet et al., 2011). The control of these

1 This work was supported by a Discovery Grant from DuPontPioneer.

* Address correspondence to [email protected] author responsible for distribution of materials integral to the

findings presented in this article in accordance with the policy de-scribed in the Instructions for Authors (www.plantphysiol.org) is:Bailin Li ([email protected]).

S.R.T., O.N.D., J.H., and B.L. conceived and designed the experi-ments; S.R.T., X.M., M.B., G.Z.-H., C.H., and B.V.A. performed theexperiments and contributed to data analysis; S.R.T. and B.L. wrotethe article.

[OPEN] Articles can be viewed without a subscription.www.plantphysiol.org/cgi/doi/10.1104/pp.15.01267

586 Plant Physiology�, January 2016, Vol. 170, pp. 586–599, www.plantphysiol.org � 2016 American Society of Plant Biologists. All Rights Reserved. www.plantphysiol.orgon March 25, 2020 - Published by Downloaded from

Copyright © 2016 American Society of Plant Biologists. All rights reserved.

http://orcid.org/0000-0002-3996-6617

http://orcid.org/0000-0002-0092-6845

mailto:[email protected]

http://www.plantphysiol.org

mailto:[email protected]

http://www.plantphysiol.org/cgi/doi/10.1104/pp.15.01267


alternative splicing differences among different tissues isthought to be the result of differential splicing factorexpression and is likely also influenced by tissue-specificmethylation patterns (Regulski et al., 2013). Alternativesplicing has also been tied closely with nonsense-mediated decay (NMD) and has been shown to in-crease or decrease a transcript’s sensitivity to this decaymechanism under a variety of conditions (Kalyna et al.,2012; Drechsel et al., 2013; Staiger and Brown, 2013).Plants possess a wide array of regulatory mechanisms

to deal with environmental stresses, which act at boththe transcriptional and posttranscriptional levels. Alter-native splicing has recently been implicated as playing amajor role in these processes during stresses, includingdisease, heat, drought, and others (Deom et al., 1987; Iidaet al., 2004; Jeong et al., 2011; Kakumanu et al., 2012;Chang et al., 2014; Mandadi and Scholthof, 2015). Geneexpression changes in splicing factors are a major fac-tor in determining stress-induced alternative splicingchanges (Yan et al., 2012; Leviatan et al., 2013; Staigerand Brown, 2013). For example, Cui et al. (2014) recentlyshowed that overexpression of the splicing factor SAD1conferred increased splicing precision and resulted inincreased tolerance to salt stress inArabidopsis (Arabidopsisthaliana). These and other findings highlight the impor-tance of examining mRNA changes at both the tran-scriptional and posttranscriptional levels to obtain amore complete picture of both development and stress.In order to assess the role of alternative splicing

during drought and development, we constructed andanalyzed 94 Illumina RNA-seq libraries from B73developing leaf, ear, and tassel under well-watered anddrought conditions. These data were supplementedwith 53 publicly available maize B73 libraries fromdeveloping seed, endosperm, and embryo tissues(Chen et al., 2014). This analysis uncovered more than48,000 novel isoforms of known genes and led to theidentification of thousands of developmentally regu-lated splicing changes, many of which were affected bydrought stress. These changes mostly occurred in leafand ear, where drought caused a large increase in de-velopmental splicing changes, while tassel was nearlyunchanged by stress. Analysis of gene expressionchanges in genes associated with alternative splicinguncovered a linear relationship between the number ofsplicing factors with gene expression changes and thenumber of alternative splicing events in other genes

during development. Together, these results highlightthe importance of alternative splicing in developingmaize vegetative and reproductive tissues.

RESULTS AND DISCUSSION

Discovery of Novel Transcripts

To examine the regulation of alternative splicingduring development under well-watered and droughtfield conditions, a series of 94 RNA-seq libraries wasconstructed and analyzed. This series included ear,tassel, and leaf tissue at four developmental stages(V12, V14, V18, and R1) under both well-watered anddrought conditions, with four biological replicateseach. After processing and genome matching (Kimet al., 2013), a total of over 1.3 billion genome-matchedreads were utilized in downstream analysis (Table I).This data analysis was also supplemented with a pub-licly available time series of developing seed, endo-sperm, and embryo (Chen et al., 2014), which resultedin another 800 million genome-matched reads (Table I).

Genome-matched reads from all libraries were ana-lyzed via Cufflinks (Trapnell et al., 2010) in order to as-semble novel transcripts, which may have been missedpreviously due to tissue- or stage-specific expression.Novel transcripts were then filtered to remove low-abundance isoforms, which could be the result of se-quencing processing intermediates or incorrect transcriptassembly. Transcripts were required to have an averageexpression among four biological replicates of at least 1.3fragments per kilobase per million reads (FPKM) in atleast one stage and condition for ear, tassel, and leaf inorder to optimize the tradeoff between false-positive andfalse-negative discoveries (Thatcher et al., 2014). Sincemany seed, endosperm, and embryo stages lacked bio-logical replicates, the requirement for these tissues was anaverage of 1.3 FPKM in at least four biological replicatesand/or consecutive developmental stages in at least onetissue. After abundance filtering, a total of 48,432 noveltranscripts were identified (Supplemental Table S1).

Broad Tissue- and Stage-Specific Isoform Switches

In order identify developmentally regulated changesin alternative splicing, Cuffdiff was used to quantify

Table I. Summary statistics for RNA-seq libraries

Description Condition Stages Libraries Total Reads Genome Matched Uniquely Mapped Percentage Mapped

Ear Watered 4 16 286,248,214 261,827,251 233,864,244 91Tassel Watered 4 16 215,381,271 205,082,097 180,285,669 95Leaf Watered 4 16 230,210,634 204,178,390 183,473,272 89Ear Drought 4 16 299,932,575 256,161,636 229,619,752 86Tassel Drought 4 14 207,741,601 197,232,548 173,040,626 95Leaf Drought 4 16 241,874,221 224,097,862 202,057,363 93Seed Watered 21 28 381,174,028 336,649,861 309,166,899 88Endosperm Watered 17 21 332,421,896 300,859,239 267,102,578 91Embryo Watered 15 16 217,398,766 197,532,915 189,422,320 91

Plant Physiol. Vol. 170, 2016 587

Splicing during Maize Development and Drought Stress

www.plantphysiol.orgon March 25, 2020 - Published by Downloaded from Copyright © 2016 American Society of Plant Biologists. All rights reserved.

http://www.plantphysiol.org/cgi/content/full/pp.15.01267/DC1


transcript expression across all tissues, stages, andconditions (Trapnell et al., 2010). Each transcript’sFPKMabundance was then converted into a percentageof total gene expression. In order to assess the rela-tionship between the different tissues, stages, andconditions and their splicing profiles, we then per-formed principal component analysis (PCA). Isoformpercentages from genes in all samples were utilized forPCA, with a relaxed cutoff of 1 FPKM total gene ex-pression used for isoform quantification compared withisoform discovery. This initial analysis predominantlyseparated samples by tissue but lacked the resolutionto easily separate stages and conditions (SupplementalFig. S1). Two additional analyses were then carried oututilizing seed data (seed, endosperm, and embryo) anddrought stress data (well-watered and drought-stressedear, leaf, and tassel) separately. The analysis of seed dataclustered samples by tissue type, with seed samplestypically falling in between endosperm and embryo (Fig.1A). Later stages of development of embryo and endo-sperm tended to cluster farther apart from each otherthan early stages, in keeping with their divergent de-velopmental programming. These results are largelyconsistent with an earlier analysis of the same data at thegene expression level carried out by others, implyingthat splicing follows a similar pattern to gene expressionin the developing maize kernel (Chen et al., 2014).

Samples included in the analysis of leaf, ear, andtassel data separate more clearly based on tissue typethan those found in seed (Fig. 1B). Consistent withthe distinct expression profiles of reproductive versusvegetative tissue, leaf clusters the farthest away fromear and tassel. Ear and tassel splicing profiles are highlysimilar at early developmental stages but diverge asdevelopment progresses, a pattern that fits well withthe morphological similarities of these two tissuesduring early development compared with later stages(Cheng et al., 1983). Under drought conditions, tasselremains largely unchanged, with well-watered anddrought-stressed samples overlaid nearly exactly ontop of each other (Fig. 1B). This lack of response at thelevel of splicing is in keeping with the lack of strongphenotypic changes to tassel under drought (Herreroand Johnson, 1981). Ear growth is known to be stronglyreduced under drought conditions (Boyer andWestgate,2004), and large splicing differences based on droughttreatment were observed (Fig. 1B). Leaf samples weretaken from the middle of the developing leaf and un-derwent comparatively fewer morphological changesover normal development and also showed few devel-opmental splicing differences. However, later stages ofdrought (V18 and R1) clustered apart from all other leafsamples, indicating an increased splicing response asdrought progressed (Fig. 1B).

In order to examine developmental splicing changesin more detail, regression analysis was carried out onthe relative isoform percentages of all genes above1 FPKM across all developmental stages of a given tis-sue. In order to be considered to have a significanttrend during development, an isoform’s relative gene

representation was required to change by at least 30percentage points (e.g. representing 10% of total geneexpression at the first stage and 40% at the last) and itsexpression pattern was required to fit a linear or ex-ponential model at r2 $ 0.65. The r2 cutoff of 0.65 waschosen by running regression analysis on correctly or-dered developmental series and comparing the numberof trends found with the number found using thesame analysis with randomized developmental or-der at r2 cutoffs ranging from 0.1 to 0.95. At r2 of 0.65,the average false-positive rate dropped below 5%(Supplemental Fig. S2, A and B). Potential overlap be-tween tissues or conditions was determined by fittingexpression values from one condition onto significanttrend lines of another, with a significance cutoff of r2 $0.3. This relaxed cutoff was chosen to capture anypotential overlap in order to reduce the chance offalsely assigning tissue specificity to a splicing change.Despite this lower cutoff, the average isoform change thatoverlapped between two tissues fit with an r2 of 0.6. Iso-form expression information for significantly regulated

Figure 1. PCA of relative isoform percentages for all genes. A, Endo-sperm and embryo show clear separation, with seed falling in between.Stage numbers represent d after pollination (DAP). B, Leaf, ear, andtassel are clearly separated, with leaf possessing the most distinct iso-form profile. Stages are shown from vegetative stage 12 (V12) throughreproductive stage 1 (R1). Drought-stressed tissues are clearly separablefor some stages of ear and leaf but indistinguishable for tassel.

588 Plant Physiol. Vol. 170, 2016

Thatcher et al.






genes in all tissues and conditions can be found inSupplemental Table S2.

Predominant Developmental Splicing Types AreTissue Specific

Next, we categorized the differences between theisoform that lost the most relative gene representationand the isoform that gained the most relative generepresentation over the course of development for eachsignificant gene. This pairwise comparison revealedthat roughly half of the developmental isoform changesare the result of alternative 59 start sites or modified39 tails (Supplemental Table S3), with many isoformchanges composed of multiple splicing and start sitemodifications (Table II). Alternative donor/acceptorchanges represent around 40% of the total develop-mental splicing changes, with little variation betweentissue types (Table II). Exon skipping, on the otherhand, is comparatively rare among most tissue types(5% of all events) except in developing ears, where itrepresents more than 27% of all events under both well-watered and drought conditions. Exon addition issimilarly rare in most tissues (6% of all events) butdominates developmental splicing in drought-stressedleaves, where it represents over 45% of all events. Intronretention/excision is the second most common differ-ence in most tissues, but it is much rarer in developingears and leaves. Overall, the tissue-specific nature ofthese different developmental splicing biases impliesthat splicing changes are likely controlled by changes indifferent splicing factors in different tissues.

NMD Sensitivity Is Affected by Isoform Switching duringDevelopment and Drought

Alternative splicing and transcription start sites havethe potential to alter a transcript’s sensitivity to NMD(Rayson et al., 2012). Although the rules governing

which transcripts are sent through the NMD pathwayremain to be fully elucidated, recent studies suggestthat three major factors in plants include 39 untrans-lated regions (UTRs) longer than 350 nucleotides, exonswithin 39 UTRs, and open reading frames greater than35 amino acids within 59 UTRs (Kalyna et al., 2012). Toassess the possibility that isoform switching interactswith the NMD pathway during development anddrought, the isoform that lost the most relative generepresentation was compared with the isoform thatgained the most relative gene representation for allgenes with significant isoform switching duringdrought or development. Overall, about half of allisoform switches were found to result in a change in thedominant isoform’s likelihood of being targeted forNMD (Table III). This ratio is relatively constant acrossall tissues and conditions. Furthermore, no obvious biaswas found toward a general decrease or general in-crease in NMD over the course of development ordrought stress. Together, these findings suggest thatNMD may play a major role in around half of all in-dividual development- and stress-specific isoform-switching events, but neither development nordrought resulted in an overall increase or decrease insplicing-related NMD changes.

Seed, Embryo, and Endosperm Undergo Hundreds ofDevelopmental Isoform Changes

Seed, embryo, and endosperm had 15 to 21 devel-opmental stages, and trends were required to contain atleast 10 of these stages. Seed, which contained amixtureof many different tissues, had the most significant iso-form trends (687), followed by endosperm (483) andembryo (369). In total, 43% of the significant transcriptchanges from endosperm and 49% from embryo over-lapped with those found in seed, indicating a gooddegree of reproducibility despite seed containing amixture of different tissues (Fig. 2A). In contrast, only100 of the 752 trends found in endosperm and embryo

Table II. Isoform switch categories by tissue type and condition

CategoryAlternative

Acceptor

Alternative

Donor

Exon

Addition

Exon

Skipping

Intron

Excision

Intron

Retention

Alternative

Start

Alternative

Tail

Seed 65 51 36 30 81 62 309 337Embryo 34 24 14 15 42 38 138 155Endosperm 42 41 13 18 52 27 200 206Leaf, watered 2 1 0 0 1 2 13 9Leaf, drought 18 13 37 1 4 9 70 57Ear, watered 24 16 10 31 14 16 116 101Ear, drought 199 140 35 210 100 83 1,008 749Tassel, watered 91 80 48 32 125 50 408 446Tassel, drought 101 77 53 43 133 57 400 418Leaf, drought stage independent 188 129 146 47 62 93 386 409Ear, drought stage independent 169 105 80 54 62 78 323 351Tassel, drought stage independent 1 2 0 0 5 3 1 7







were shared by the two tissues, while 67 out of the 1,149developmentally regulated transcripts were found inall three tissue types (Fig. 2A). The low degree ofoverlap between endosperm and embryo indicates ahighly tissue-specific regulation of developmental iso-form switching, especially when compared with themuch higher overlap with seed. The overall largeramount of significant trends found only in seed may bethe result of it containing multiple tissue types, but itcould also be influenced by having the most stagespresent in our analysis (Table I). To assess this possi-bility, the same analysis was run utilizing only the final15 stages of seed, which overlapped with embryo andendosperm. This smaller data set reduced the numberof seed-only significant trends from 397 to 230, bringingit in line with endosperm and embryo.

In order to estimate the false-positive rate of signifi-cant trends for each tissue, the same regression analysiswas rerun in all three tissues with stages placed out ofthe correct developmental order. For example, seedsamples were placed in the order 21, 10, 20, and 9 DAP,etc. This disordered data set yielded dramatically re-duced significant trends, with seed, endosperm, andembryo yielding 33, 14, and 11 significant trends, re-spectively. The large reduction in significant changes inan incorrectly ordered time series indicates that thefindings in the correctly ordered set are of high confi-dence and likely have a false-positive rate of around 3%to 5%. This false-positive rate is also consistent with the5% error rate that has been predicted when selecting anr2 of 0.65 for significant trends in a regression analysis(Bryhn and Dimberg, 2011).

Although significant changes in splicing take placeacross all developmental stages, it is possible that cer-tain stages may have higher or lower average rates ofchange. To asses this possibility, the SD of all signifi-cantly regulated isoforms within overlapping three-stage windows was calculated. Variability was seen inall tissues, with the most consistent three-stage win-dows having an average SD of 4.9 percentage points,while others had up to 7.7 percentage points. All threetissues showed a virtually identical pattern of average

isoform variability (Fig. 2B). Valleys representing pe-riods of relative consistency around 3, 18, and 34 DAPand peaks with much higher rates of change around8 and 28 DAP were present at similar levels in all threetissues (Fig. 2B). This highly consistent pattern of iso-form variability is noteworthy given the relatively lowlevel of overlap of individual transcripts (Fig. 2A) andindicates that shifts in isoforms may be controlled byone or a few master regulatory networks, such aschanges in hormone levels. This is also consistent withan earlier dendrogram analysis of gene expression inthe developing seed, which clustered stages into early,middle, and late development with transition points at8 and 28 DAP (Chen et al., 2014). Taken together, theseresults indicate that gene expression and isoformswitches are highly correlated in the developing maizeseed and display similar periods of consistency andvariability.

A small number of genes contain significant isoformswitches in all three tissue types (Fig. 2A). This group isexemplified by GRMZM2G104658, a kinase with un-known function. GRMZM2G104658 is highly alter-natively spliced and makes nine different knowntranscripts. Over the course of development, all threetissue types gradually reduced isoform 1 and increasedisoform 7. This change resulted in a switch from aprotein encoding only the kinase domain to one thatalso includes a small ligand-binding site that typ-ically has regulatory function, which implies thatGRMZM2G104658 may play a previously unknownrole in the developing seed (Fig. 3A).

Although the majority of significant changes in iso-forms were tissue specific, the functional classificationof genes encoding them frequently overlapped. Severalgenes involved in hormone synthesis or response weresignificantly regulated, including NAC TRANSCRIP-TION FACTOR109 (NACTF109; GRMZM2G014653).NACTF109 is known to produce five different tran-scripts, which mostly result in alterations of theprotein’s C-terminal domain. A novel transcript wasidentified in embryo, which also codes for a proteinwith an altered C terminus immediately following the

Table III. NMD changes by tissue type and condition

CategorySame

NMD

Higher

NMD

Lower

NMD

39 ExonAddition

39 ExonLoss

Long 39Addition

Long 39Loss

59 Upstream

Open Reading

Frame Addition

59 Upstream Open

Reading Frame Loss

Seed 198 135 126 68 61 77 80 76 61Embryo 84 84 40 52 22 56 20 43 37Endosperm 111 87 93 35 54 57 39 47 55Leaf, watered 11 3 4 1 3 2 3 1 1Leaf, drought 44 17 24 14 6 4 11 11 21Ear, watered 65 51 33 17 16 25 18 37 21Ear, drought 591 362 225 66 143 230 102 248 142Tassel, watered 249 165 156 77 75 103 106 94 78Tassel, drought 233 155 172 82 81 97 115 89 90Leaf, drought stage independent 304 205 190 111 96 114 110 100 93Ear, drought stage independent 284 156 172 89 97 96 93 83 92Tassel, drought stage independent 5 2 1 2 1 3 1 0 0


Thatcher et al.



highly conserved NAC transcription factor domain(Fig. 3B). Over the course of development, embryogradually switched from producing known isoform 3to this novel isoform with an altered tail (Fig. 3B).NACTF109’s Arabidopsis homolog, ANAC2, regulatesABA biosynthesis and has been implicated in both de-velopment and stress regulation (Wu et al., 2009; Jensenet al., 2013; Garapati et al., 2015). The alternativesplicing change in the maize NACTF109 C-terminalregion may result in altered protein localization or ac-tivity and may also affect its NMD targeting due to achange in 39UTR length. This, in turn, could affect ABAlevels in the developing embryo, where it is knownto control dormancy and other processes (Rivin andGrudt, 1991).A number of genes associated with disease respon-

siveness were also regulated over the course of devel-opment, predominantly in endosperm. This group isexemplified by GRMZM2G028568, a homolog of theArabidopsis wall-associated kinase disease responsefactor LEAF RUST10 DISEASE-RESISTANCE LOCUSRECEPTOR-LIKE PROTEINKINASE-LIKE (LRK10L). Anovel maize LRK10L isoform with an alternative tran-scription start site and splicing pattern was identifiedduring transcript discovery. This isoform lacks the ex-tracellular galacturonan-binding wall-associated ki-nase (GUB WAK) domain and codes for a differentdownstream Cys-rich WAK domain (Fig. 3C). Over the

course of development, the novel isoform graduallydeclines and known isoform 1 comes to dominateLRK10L expression (Fig. 3C). The regulation of this andother disease resistance genes over the course of en-dosperm development may be the result of increasingpressure from pathogens as themaize kernel transitionsthrough its grain-filling stage and increases its carbo-hydrate content, or it could point to previously un-known developmental roles for these genes (Myerset al., 2000).

Several genes associated with creating or reading epi-genetic marks were found to be alternatively spliced overdevelopment in all three tissues. This group is representedby the methylation-sensitive DDT-TRANSCRIPTIONFACTOR6 (DDT6; GRMZM2G314661). Two novel iso-forms ofDDT6were identified, which differ by a minorchange in a splice site acceptor and a small additionalintron removal (Fig. 3D). The splice site acceptormodification results in the gain of a histone-interactingWHIM domain in novel isoform 2 compared with iso-form 1, but the downstream intron removal causes theloss of a plant homeodomain fold involved in methyl-ated histone H3 binding in this same isoform (Fig. 3D).Additionally, the second novel isoform has a dramati-cally increased 39UTRwith multiple exons, making it alikely target for NMD (Fig. 3D). Early in developmentof the seed, isoform 2 dominates DDT6 expression, butthe two isoforms gradually come to roughly equal ex-pression levels by 21 DAP (Fig. 3D). Epigenetic markshave been associated previously with tissue identity(Luo et al., 2009), and here, we show that genes in-volved in epigenetic pathways are dynamically regu-lated at the level of alternative splicing as developmentprogresses in the seed.

Ear, Tassel, and Leaf Have Thousands of DevelopmentallyRegulated Isoform Changes under Well-Watered andDrought Conditions

Next, we examined significant developmental trendsin leaf, ear, and tassel under both well-watered anddrought conditions. Under well-watered conditions, atotal of 1,204 isoforms had significant developmentalregulation, while 3,064 were found during drought.Similar to the analysis conducted on seed samples, wedetermined the reliability of this data set by testing itwith a disordered set: V18, V12, R1, and V14. Thisdisordered data yielded only 40 and 22 significantchanges under well-watered and drought conditions,respectively. The dramatic reduction in significantchanges in a disordered time series again indicated thatthe findings in the correctly ordered set are likely tohave a false-positive rate of around 2% to 4%.

In order to determine the overlap of developmentallyregulated transcripts between different tissues, datafrom one condition were fitted onto significant trendlines generated from regression analysis of the othercondition, with a relaxed criterion of r2 $ 0.3 (Fig. 4A).Analysis of tissue-level overlap showed that the vast

Figure 2. Alternative splicing changes over the course of seed, endo-sperm, and embryo development. A, Significant isoform changes overthe course of development in seed, endosperm, and embryo. Transcriptsare considered significant if their relative gene representation changedby at least 30 percentage points over the course of development and fit alinear or exponential regression at r2 $ 0.65. B, Average SD of all sig-nificantly regulated isoforms within overlapping three-stage windowsshowing that seed, endosperm, and embryo all have similar stage-specific isoform variation rates.





majority of developmental changes were tissue specific(Fig. 4A). Under well-watered conditions, very littleoverlap was seen between ear and tassel (19 tran-scripts), and leaf did not share any developmentalregulation with either other tissue. Under drought, ear

increased its overlap with tassel (34 transcripts) andgained a low level of overlap with leaf (eight tran-scripts). Overall, this low level of overlap indicateda high degree of tissue specificity in developmentalisoform switches.

Figure 3. Examples of genes with developmentally regulated alternative splicing in embryo, seed, and endosperm. The two mostabundant transcripts quantified by RNA-seq for each gene are shown. For transcript models, black denotes a known transcriptwhile white is a novel transcript assembled by Cufflinks. Start and stop coordinates for the longest predicted proteins are markedby gray arrows. For protein models, conserved domains that are lost or gained are denoted in red while consistent domains aredepicted in gray. A, KinaseGRMZM2G104658 switches between a proteinwith or without a small ligand-binding ACT domain inembryo, endosperm, and seed development. B, The abscisic acid (ABA) synthesis regulatorNACTF109 switches from producing aprotein with a longer C-terminal tail to a shorter tail over embryo development. C, Disease resistance-related LRK10L switchesfrom producing a protein without a GUBWAK domain to one with the GUBWAKover the course of endosperm development. D,Methylation-sensitive transcription factorDDT6 exchanges an additional plant homeodomain (PHD) for aWHIM domain duringseed development.


Thatcher et al.



In keeping with the earlier PCA results (Fig. 1), leafunderwent the smallest total number of isoformswitches over the course of development, with only 15isoforms showing a significant change under well-watered conditions. Only two of these changes werealso found in drought at r2 $ 0.3, suggesting thatdrought causes a large deviation from normal devel-opmental programming (Fig. 4B). This difference can beclearly seen in CORONATINE INSENSITIVE1 (COI1;GRMZM2G353209), a jasmonate-responsive compo-nent of a ubiquitination complex involved in defenseand drought responses (Xie et al., 1998; Harb et al.,2010). Transcript discovery revealed a novel isoform ofCOI1 that codes for a similar protein that retains acrucial F-box, shown to be crucial for ubiquitinationtarget selection in Arabidopsis, but lacks a Leu-richrepeat region (Fig. 5A; Xu et al., 2002). Under well-watered conditions, COI1 gradually changes fromproducing the known full-length isoform to produc-tion of the novel isoform (Fig. 5A). Under drought

conditions, COI1 continues to produce the known iso-form containing the Leu-rich repeat region, potentiallyresulting in increased jasmonate sensitivity in olderleaves under drought compared with well-wateredconditions.

Substantially more genes with isoform switching overthe course of development were found when leaveswere stressed with drought than under well-wateredconditions (Fig. 4B). This group included severalgenes associated with stress response, such asHEATSHOCK PROTEIN93-V (HSP93-V; GRMZM2G009443),a chloroplast-associated heat shock protein. A novelisoform ofHSP93-V that lacks both Clp protein-bindingsites was identified (Fig. 5B). Under well-wateredconditions, this truncated isoform predominates,but drought causes a gradual increase in the full-length known transcript containing these Clp do-mains (Fig. 5B). Clp domains are crucial for thefunction of this class of heat shock protein, and theswitch to an isoform that codes for them underdrought stress conditions could function to increaseprotein degradation or prevent protein aggregation(Wang et al., 2004).

Ear showed significantly more developmental regu-lation of isoform switching under well-watered condi-tions compared with leaf (119 isoforms), and themajority of this regulation (73 isoforms) was consistentdespite drought conditions. Still, large deviationsfrom normal developmental changes were seen insome genes, such as MONODEHYDROASCORBATEREDUCTASE4 (MDAR4; GRMZM2G320307). MDAR4is a membrane-bound monodehydroascorbate reduc-tase that functions in the ascorbate-glutathione cycleto remove toxic hydrogen peroxide, but it also hasbeen implicated in seed storage and postgerminationgrowth in Arabidopsis (Chew et al., 2003; Eastmond,2007). Under well-watered conditions, known isoform1 slowly reduces over the course of development,while a second known isoformwith a longer C-terminaltail steadily increases. This isoform also has a muchshorter 39 UTR, which could potentially alter the tran-script’s sensitivity to NMD (Kalyna et al., 2012; Fig. 5C).Under drought conditions, isoform 1 continues todominate the gene expression from V12 through R1.These changes in MDAR4 may be a response to stress-induced increases in hydrogen peroxide but also couldaffect seed storage as development progresses to laterstages.

During drought, ear showed substantially moredevelopmental regulation of isoform switching(1,512 isoforms; Fig. 4B). In keeping with this largeincrease in alternative splicing during drought,several splicing factors were identified as droughtregulated, such as the salicylic acid-responsive RNA-BINDING PROTEIN-DEFENSE RELATED1 (BRN1;GRMZM2G005459). BRN1 was found to produce anovel isoform that codes for a protein with two, ratherthan three, RNA-binding domains (Fig. 5D). Underwell-watered conditions, this slightly truncated iso-form gradually came to dominate BRN1’s total gene

Figure 4. Significant isoform and gene expression changes over thecourse of development and drought in leaf, ear, and tassel. Changes intranscripts were considered significant if their relative gene represen-tation changed by at least 30 percentage points over the course of de-velopment and fit a linear or exponential regression at r2 $ 0.65. A,Overlap between developmental isoform switches in different tissues.B, Overlap between developmental isoform switches under well-watered and drought conditions. C, Overlap of developmental geneexpression changes between well-watered and drought conditions inleaf, ear, and tassel.





Figure 5. Examples of genes with developmentally regulated alternative splicing under well-watered and drought conditions inleaf, ear, and tassel. The two most abundant transcripts quantified by RNA-seq for each gene are shown. Error bars represent SE

from four biological replicates. For transcript models, black denotes a known transcript and white is a novel transcript assembledby Cufflinks. Start and stop coordinates for the longest predicted proteins are marked by gray arrows. For protein models, con-served domains that are lost or gained are denoted in red while consistent domains are depicted in gray. A, The splicing pattern ofthe jasmonate-responsive ubiquitination component COI1 is changed during development in leaf only under well-wateredconditions. B, The heat shock protein HSP93-V undergoes alternative splicing changes in leaf only under drought. C, MDAR4undergoes alternative splicing changes in ear under well-watered conditions. D, The splicing factor BRN1 undergoes alternativesplicing changes in ear only under well-watered conditions. E, The splicing factor U2AF65A undergoes alternative splicingchanges in tassel under drought.


Thatcher et al.



expression, potentially significantly affecting thesplicing ofmany other genes (Fig. 5D). The regulation ofthis defense-related splicing factor in drought-stressedears may represent a splicing-related response to in-creased disease pressure under more stressful con-ditions (Qi et al., 2010).Tassel had the most developmental regulation under

well-watered conditions (914 isoforms), and almost 80%(716 isoforms) of these maintained a similar patternunder drought conditions. This high degree of similarityin developmental splicing under well-watered anddrought conditions was also in keeping with earlierPCA results (Fig. 1). Despite this, some genes didshow significant differences under drought condi-tions. Splicing factor U2 small nuclear ribonucleo-protein auxiliary factor (U2AF65A; GRMZM5G813627)is known to produce a number of different func-tional isoforms, with alteration in its three RNArecognition motifs (RRMs). Under well-wateredconditions, isoform 1, which contains RRMs 1 and2, is always the dominant isoform. Under droughtconditions, however, isoform 2 (which contains allthree RRMs) dominates expression early on andgradually diminishes until, at R1, U2AF65A’s splic-ing pattern under drought becomes nearly identicalto well-watered conditions (Fig. 5E). Another U2AFsplicing factor, ROUGH ENDOSPERM3, has beenimplicated in seed development (Fouquet et al.,2011), while GRMZM5G813627 may instead play arole in tassel.

Developmentally Consistent Drought-InducedIsoform Changes

Next, we analyzed drought-induced isoform changesthat occurred in a developmental stage-independentmanner. Isoform percentages for genes with greaterthan 1 FPKM expression in at least three out of fourdevelopmental stages under both well-watered anddrought conditions were compared with each other.Isoform abundances with an average of at least a30 percentage point difference between well-wateredand drought conditions were then compared betweenconditions utilizing Student’s t test (critical value0.999). Thousands of consistent changes were seen,including 1,060 in leaf, 932 in ear, but only 11 in tassel(Fig. 6A; Supplemental Table S4). Tassel’s low num-ber of consistent isoform changes under drought wassimilar to its small changes in developmentally con-trolled isoform switching under drought (Fig. 6B).Stage-independent drought isoform changes in leafand ear, on the other hand, were significantly differ-ent. Leaf had the smallest number of developmentalchanges under both drought and well-watered con-ditions but showed the most stage-independent iso-form changes during drought (Figs. 4B and 6A). Incontrast to developmentally controlled drought re-sponses, where ear and leaf had very little overlap,developmental stage-independent changes were

highly similar between these tissues (Figs. 4A and 6A).These findings indicate that stage-specific drought al-ternative splicing tends to be highly tissue specific,while stage-independent drought changes occur in anumber of tissues simultaneously.

A comparison of the types of alternative splicingpresent in leaf and ear under drought conditionsrevealed a fairly similar distribution of categories tothat seen in developmental splicing (Tables II andIII). Alternative donor/acceptor dominated the alter-native splicing seen in leaf and ear to a greater degree.Unlike developmental splicing changes, drought-inducedsplicing showed no clear bias toward skipping or

Figure 6. Significant isoform switches between well-watered anddrought conditions that are consistent across developmental stages.A, Tissue overlap of significant drought-induced alternative splicingchanges that were consistent across at least three out of four develop-mental stages. B, DCD (GRMZM2G151440) gene expression is unaf-fected by drought but undergoes significant changes in alternativesplicing in leaf and ear. Gene expression comparison between samplesis shown in blue. Relative gene representation of the two isoformswithin a sample is shown in red. C,DCD splicing changes affect highlyconserved amino acid residues (arrowheads) within the catalytic do-main (gray boxes). Amino acid residues and protein regions that arealtered during the isoform switch are denoted in red. Start and stopcoordinates for the longest predicted protein aremarked by gray arrows.






addition. Overall, stage-independent drought splicingcategories were much more similar in leaf and ear thanchanges seen during their development, which wasconsistent with the much higher overlap of individualtranscripts in the stage-independent splicing set (Fig.6A). Tassel had very few stage-independent droughtsplicing events, but eight out of 11 of these involved achange in intron retention/excision, which was also thepredominant category seen during tassel developmentunder well-watered and drought conditions (Tables IIand III).

Many stage-independent changes looked virtuallyidentical in leaf and ear but were completely absent intassel, a trend exemplified by D-CYSTEINE DESULF-HYDRASE (DCD; GRMZM2G151440). Under droughtconditions, DCD undergoes a consistent and dramaticswitch from a novel slightly truncated isoform to itsknown full-length isoform in both leaf and ear, but it isunchanged in tassel (Fig. 6B). The known isoform,which is induced under drought, contains the fullsubstrate-binding pocket, compared with the novelisoform, which dominates under well-watered condi-tions (Fig. 6C). Interestingly, DCD’s Arabidopsis ho-molog has been shown to be induced during droughtstress and is thought to confer drought tolerancethrough increased production of hydrogen sulfide (Jinet al., 2011). Maize DCD is not regulated at the geneexpression level in our data but instead undergoes asplicing change that increases the ratio of fully activeprotein, potentially causing increased hydrogen sulfideproduction through alternative splicing instead of geneexpression (Fig. 6, B and C).

Changes in Splicing Correlate with Gene ExpressionChanges in Splicing Factors

Next, we analyzed gene expression over the course ofdevelopment. Genes were considered significantlyregulated if they hadmore than a 3-fold change and fit alinear or exponential regression model at r2 $ 0.65.

Although the total number of genes with significantexpression changeswas greater than thosewith isoformchanges, the pattern of both tissue- and condition-specific regulation was largely similar to that seen inthe alternative splicing data (Fig. 4). One exception tothis similar pattern was drought-stressed ears, whichunderwent a large induction in developmental iso-form switching but only a modest change in devel-opmental gene expression changes during drought(Fig. 4, B and C). Additionally, although seed under-went the largest number of developmental isoformswitches, endosperm underwent the largest numberof developmental gene expression changes (Fig. 2A;Supplemental Fig. S3).

A more focused analysis was then carried out on theexpression change of genes associated with alternativesplicing regulation based on Gene Ontology (GO). Atotal of 293 maize genes that had GO terms associatedwith splicing or had orthologs in Arabidopsis or rice(Oryza sativa) associated with splicing were examined.A total of 77 known or putative splicing factors werefound to be regulated in at least one tissue andcondition, with 31% occurring in more than onetissue (Fig. 7A; Supplemental Table S5). A similarcomparison of splicing changes across all six tissuesyielded even more tissue specificity, with only 10% ofall splicing changes occurring in more than one tissue(Fig. 7B). The difference in specificity of splicing fac-tor gene expression changes compared with splicingchanges likely arises due to the influence of otherfactors on splicing, such as transcription rate ormethylation status (Kornblihtt et al., 2004; Shepardand Hertel, 2008; Regulski et al., 2013). It has beenshown previously that splicing factors are fre-quently alternatively spliced, often resulting in in-creased or decreased NMD sensitivity (Lareau andBrenner, 2015). In total, we identified 77 splicingfactors with gene expression change (SupplementalTable S5) and 46 splicing factors that are alternativelyspliced. However, only six of these potential splicingfactors were subject to both layers of regulation during

Figure 7. Significant expression changesin genes associated with splicing andsignificant alternative splicing changes inall genes. A, Tissue overlap of genes withGO terms associated with splicingwhose expression changes by more than3-fold and fits a linear or exponentialregression line at r2 $ 0.65. B, Tissueoverlap of splicing isoforms whose rela-tive gene representation changes bymore than 30 percentage points and fits alinear or exponential regression line atr2 $ 0.65.


Thatcher et al.







development, none of which showed signs of geneexpression regulation via NMD sensitivity changes(Supplemental Tables S2 and S5).When comparing across all tissues under well-

watered conditions, the number of splicing-regulatinggenes with developmental expression changes has alinear relationship with the number of genes that un-derwent alternative splicing changes over development(Fig. 8A). On average, for every splicing factor withsignificant gene expression change, 21 genes have sig-nificant splicing changes. This trend held despite theuse of disparate tissue types, with time point numbersranging from four to 21. Under drought stress, the trendno longer held across tissues (Fig. 8B). The loss of thelinear relationship of splicing factor regulation withalternative splicing under drought may be due to theeffects of transcription rate or osmotic stresses on RNAsecondary structure, both of which could potentiallyalter splicing independent of changes in the expressionlevel of splicing factors (Shepard and Hertel, 2008).Additionally, the massive increase in developmentalsplicing changes in drought-stressed earsmay be due toan increase in the number of putative splicing factorsthat are alternatively spliced from three to 24.Intriguingly, the expression level of splicing factor

PRP18 (GRMZM2G102711) seems to correlate wellwith drought-responsive alternative splicing. Underwell-watered conditions, this gene has only minorregulation over development. However, ear and leaf,which show dramatic increases in splicing underdrought conditions, show large increases in expres-sion for this gene under drought conditions (Fig. 9).

On the other hand, no expression change for thisgene was seen in tassel, which had few drought-induced developmental splicing changes (Figs. 8and 9). In Arabidopsis, overexpression of the saltstress-regulated splicing factor SAD1 was shown toincrease splicing precision and confer salt tolerance(Cui et al., 2014). PRP18, therefore, may play a sim-ilar role in drought stress response and warrantsfurther investigation.

CONCLUSION

Deep sequencing of maize leaf, ear, and tassel tissuesover development and during drought stress as well asinclusion of public data from developing maize seed,endosperm, and embryo tissues enabled the identifi-cation of thousands of developmental and stress-relatedsplicing changes. Ninety percent of these splicingchanges were specific to a single tissue, indicating atight control of developmental alternative splicingchanges. Drought was found to cause massive shifts indevelopmental splicing changes in leaf and ear butonly minor changes in tassel. Dozens of known orputative splicing factors were found to be regulated atthe gene expression level during development, and alinear relationship was found between expressionchanges in splicing factors and splicing changes inother genes. Taken together, these findings highlightthe important role that alternative splicing plays intissue-specific development and response to droughtstress in maize.

Figure 9. Splicing factor PRP18’sexpression change over the course ofdevelopment in leaf, ear, and tassel. Geneexpression forPRP18 (GRMZM2G102711)increases over development under droughtstress in both leaf and ear but is unchangedin tassel. Data were generated from RNA-seq quantification. Error bars representSE from four biological replicates.

Figure 8. Number of genes with signif-icant alternative splicing changes versusthe number of splicing-related genes withexpression changes. A, Well-watered con-ditions show a linear relationship betweenthe number of genes associated withsplicing that are under expression reg-ulation and the number of alternativelyspliced genes. B, Under drought, noclear relationship is seen betweensplicing gene regulation and alter-native splicing changes.






MATERIALS AND METHODS

Plant Growth and Harvesting

B73 maize (Zea mays) plants were grown in a field in Woodland, California,in 2012. Plots were designed utilizing a randomized complete block designwithfour biological replicates. Drought stress was applied by withholding waterafter stage V8, while well-watered plants received constant irrigation to achievemaximum yields. Tissues were harvested 11, 18, 27, and 32 d later from bothwell-watered and drought-stressed plants. Time points corresponded to stagesV12, V14, V18, and R1 for well-watered plants, with minor stage delays seen indrought-stressed plants. All tissues were harvested from the same plants. Leaftissue was cut from the middle section of the top collared leaf from three dif-ferent plants in each replicate group. Ears were selected based on size at eachtime point: approximately 0.5 cm, 0.8 to 1.5 cm, 2 to 4 cm, and 3 to 8 cm. The firsttime point was composed of 32 ears, the second 16 ears, and the third and fourththree to four ears.Whole tassels were dissected fromwhorls during the first twotime points. Tassel emergence appeared at the third time point and pollenshedding at the last. After this point, 10 cm of the main spike was sampled. Alltissues were immediately frozen in liquid nitrogen and then utilized for mRNAsequencing library preparation.

mRNA Sequencing Library Preparation and Sequencing

Total RNAwas isolated from frozenmaize tissueswith theQiagenRNeasyKitfor total RNA isolation. Libraries from total RNA were then prepared using theTruSeqmRNA-SeqKit and protocol from Illumina and sequenced on the IlluminaHiSeq 2500 systemwith Illumina TruSeq SBS version 3 reagents. On average,18million readswere generated from each library (Table I). Resulting sequenceswere trimmed based on quality scores (Phred score $ 13) and mapped to themaize B73 reference genome sequence V2 andmaize working gene set V5awithTopHat2 version 2.0.14 (Kim et al., 2013) using several modifications fromdefault parameters: maximum intron size, 100,000; minimum intron size, 20;and up to two mismatches allowed. Reads that aligned to multiple locationswere assigned heuristically based on the abundance of surrounding regions(Kim et al., 2013). Libraries with less than 5,000,000 genome-matched reads (onebiological replicate of well-watered R1 tassel and one biological replicate ofdrought-stressed R1 tassel) were excluded from later downstream analysis.

Computational Prediction of Novel Isoforms

Genome-matched reads fromeach library thenwere assembledwithCufflinksversion 2.1.1 (Trapnell et al., 2010) using several modifications from default pa-rameters:maximum intron size, 100,000; andminimum intron size, 20. Cuffmergeversion 2.1.1 (Roberts et al., 2011) was then used to merge individual transcriptassemblies into a single transcript set. The annotation of novel junctions requiredat least 10 reads spanning them and any new transcripts needed to represent atleast 10% of the total gene abundance in at least one library. Known and noveltranscripts were quantified in each tissue and genotype with Cuffnorm version2.1.1 (Roberts et al., 2011) using default parameters. Novel transcripts with ex-pression of less than 1.3 FPKM in all tissues and stages were filtered out.

Analysis of Developmental Isoform Changes

Cuffdiff version 2.1.1 (Roberts et al., 2011) was used to quantify transcriptexpression across all tissues, developmental stages, and conditions, which werethen converted to relative percentage of total gene expression. Regressionanalysis was then applied to individual transcripts to look for linear or expo-nential developmental trends, with at least a 30 percentage point change andr2 $ 0.65 required to be considered significant.

The data set supporting the results of this article is available in the NationalCenter for Biotechnology Information Gene Expression Omnibus repository(GSE71723).

Supplemental Data

The following supplemental materials are available.

Supplemental Figure S1. PCA analysis of relative isoform percentages forall tissues, stages, and conditions.

Supplemental Figure S2. Significant gene expression trends over thecourse of development.

Supplemental Figure S3. Number of isoform switches during develop-ment at various r2 cutoffs.

Supplemental Table S1. Transcripts discovered during Cufflinks tran-scriptome assembly.

Supplemental Table S2. Relative isoform percentages of total geneexpression.

Supplemental Table S3. Splicing changes versus alternative start site/tail.

Supplemental Table S4. Genes with consistent isoform switches underdrought.

Supplemental Table S5. Significant gene expression changes duringdevelopment.

ACKNOWLEDGMENTS

We thank Renee Lafitte and Salim Hakimi for managing the droughttreatment.

Received August 13, 2015; accepted November 17, 2015; published November18, 2015.

LITERATURE CITED

Boyer JS, Westgate ME (2004) Grain yields with limited water. J Exp Bot 55:2385–2394

Bryhn AC, Dimberg PH (2011) An operational definition of a statisticallymeaningful trend. PLoS ONE 6: e19241

Chang CY, Lin WD, Tu SL (2014) Genome-wide analysis of heat-sensitive alternative splicing in Physcomitrella patens. Plant Physiol165: 826–840

Chen J, Zeng B, Zhang M, Xie S, Wang G, Hauck A, Lai J (2014) Dynamictranscriptome landscape of maize embryo and endosperm development.Plant Physiol 166: 252–264

Cheng PC, Greyson RI, Walden DB (1983) Organ initiation and the de-velopment of unisexual flowers in the tassel and ear of Zea mays. Am JBot 70: 450–462

Chew O, Whelan J, Millar AH (2003) Molecular definition of the ascorbate-glutathione cycle in Arabidopsis mitochondria reveals dual targeting ofantioxidant defenses in plants. J Biol Chem 278: 46869–46877

Cui P, Zhang S, Ding F, Ali S, Xiong L (2014) Dynamic regulation ofgenome-wide pre-mRNA splicing and stress tolerance by the Sm-likeprotein LSm5 in Arabidopsis. Genome Biol 15: R1

Deom CM, Oliver MJ, Beachy RN (1987) The 30-kilodalton gene product oftobacco mosaic virus potentiates virus movement. Science 237: 389–394

Drechsel G, Kahles A, Kesarwani AK, Stauffer E, Behr J, Drewe P, Rätsch G,Wachter A (2013) Nonsense-mediated decay of alternative precursor mRNAsplicing variants is a major determinant of the Arabidopsis steady state tran-scriptome. Plant Cell 25: 3726–3742

Eastmond PJ (2007) MONODEHYROASCORBATE REDUCTASE4 is re-quired for seed storage oil hydrolysis and postgerminative growth inArabidopsis. Plant Cell 19: 1376–1387

Erkelenz S, Mueller WF, Evans MS, Busch A, Schöneweis K, Hertel KJ,Schaal H (2013) Position-dependent splicing activation and repressionby SR and hnRNP proteins rely on common mechanisms. RNA 19: 96–102

Fouquet R, Martin F, Fajardo DS, Gault CM, Gómez E, Tseung CW,Policht T, Hueros G, Settles AM (2011) Maize rough endosperm3 en-codes an RNA splicing factor required for endosperm cell differentiationand has a nonautonomous effect on embryo development. Plant Cell 23:4280–4297

Garapati P, Xue GP, Munné-Bosch S, Balazadeh S (2015) Transcriptionfactor ATAF1 in Arabidopsis promotes senescence by direct regulationof key chloroplast maintenance and senescence transcriptional cascades.Plant Physiol 168: 1122–1139

Harb A, Krishnan A, Ambavaram MM, Pereira A (2010) Molecular andphysiological analysis of drought stress in Arabidopsis reveals earlyresponses leading to acclimation in plant growth. Plant Physiol 154:1254–1271

Herrero MP, Johnson RR (1981) Drought stress and its effects on maizereproductive systems. Crop Sci 21: 105–110


Thatcher et al.











Iida K, Seki M, Sakurai T, Satou M, Akiyama K, Toyoda T, Konagaya A,Shinozaki K (2004) Genome-wide analysis of alternative pre-mRNAsplicing in Arabidopsis thaliana based on full-length cDNA sequences.Nucleic Acids Res 32: 5096–5103

Jensen MK, Lindemose S, de Masi F, Reimer JJ, Nielsen M, Perera V,Workman CT, Turck F, Grant MR, Mundy J, et al (2013) ATAF1transcription factor directly regulates abscisic acid biosynthetic geneNCED3 in Arabidopsis thaliana. FEBS Open Bio 3: 321–327

Jeong BR, Lin Y, Joe A, Guo M, Korneli C, Yang H, Wang P, Yu M, CernyRL, Staiger D, et al (2011) Structure function analysis of an ADP-ribosyltransferase type III effector and its RNA-binding target in plantimmunity. J Biol Chem 286: 43272–43281

Jin Z, Shen J, Qiao Z, Yang G, Wang R, Pei Y (2011) Hydrogen sulfideimproves drought resistance in Arabidopsis thaliana. Biochem BiophysRes Commun 414: 481–486

Kakumanu A, Ambavaram MM, Klumas C, Krishnan A, Batlang U,Myers E, Grene R, Pereira A (2012) Effects of drought on gene ex-pression in maize reproductive and leaf meristem tissue revealed byRNA-Seq. Plant Physiol 160: 846–867

Kalyna M, Simpson CG, Syed NH, Lewandowska D, Marquez Y,Kusenda B, Marshall J, Fuller J, Cardle L, McNicol J, et al (2012)Alternative splicing and nonsense-mediated decay modulate ex-pression of important regulatory genes in Arabidopsis. NucleicAcids Res 40: 2454–2469

Kim D, Pertea G, Trapnell C, Pimentel H, Kelley R, Salzberg SL (2013)TopHat2: accurate alignment of transcriptomes in the presence of in-sertions, deletions and gene fusions. Genome Biol 14: R36

Kornblihtt AR, de la Mata M, Fededa JP, Munoz MJ, Nogues G (2004)Multiple links between transcription and splicing. RNA 10: 1489–1498

Kriechbaumer V, Wang P, Hawes C, Abell BM (2012) Alternative splicingof the auxin biosynthesis gene YUCCA4 determines its subcellularcompartmentation. Plant J 70: 292–302

Lareau LF, Brenner SE (2015) Regulation of splicing factors by alternativesplicing and NMD is conserved between kingdoms yet evolutionarilyflexible. Mol Biol Evol 32: 1072–1079

Leviatan N, Alkan N, Leshkowitz D, Fluhr R (2013) Genome-wide surveyof cold stress regulated alternative splicing in Arabidopsis thaliana withtiling microarray. PLoS ONE 8: e66511

Li W, Lin WD, Ray P, Lan P, Schmidt W (2013) Genome-wide detection ofcondition-sensitive alternative splicing in Arabidopsis roots. PlantPhysiol 162: 1750–1763

Lorkovi�c ZJ, Wieczorek Kirk DA, Lambermon MH, Filipowicz W (2000)Pre-mRNA splicing in higher plants. Trends Plant Sci 5: 160–167

Luo C, Durgin BG, Watanabe N, Lam E (2009) Defining the functional net-work of epigenetic regulators in Arabidopsis thaliana. Mol Plant 2: 661–674

Mandadi KK, Scholthof KB (2015) Genome-wide analysis of alternativesplicing landscapes modulated during plant-virus interactions in Bra-chypodium distachyon. Plant Cell 27: 71–85

Myers AM, Morell MK, James MG, Ball SG (2000) Recent progress towardunderstanding biosynthesis of the amylopectin crystal. Plant Physiol122: 989–997

Pertea M, Mount SM, Salzberg SL (2007) A computational survey ofcandidate exonic splicing enhancer motifs in the model plant Arabi-dopsis thaliana. BMC Bioinformatics 8: 159

Qi Y, Tsuda K, Joe A, Sato M, Nguyen V, Glazebrook J, Alfano JR, CohenJD, Katagiri F (2010) A putative RNA-binding protein positively regu-lates salicylic acid-mediated immunity in Arabidopsis. Mol Plant Mi-crobe Interact 23: 1573–1583

Rayson S, Arciga-Reyes L, Wootton L, De Torres Zabala M, Truman W,Graham N, Grant M, Davies B (2012) A role for nonsense-mediatedmRNA decay in plants: pathogen responses are induced in Arabi-dopsis thaliana NMD mutants. PLoS ONE 7: e31917

Regulski M, Lu Z, Kendall J, Donoghue MT, Reinders J, Llaca V,Deschamps S, Smith A, Levy D, McCombie WR, et al (2013) Themaize methylome influences mRNA splice sites and reveals widespreadparamutation-like switches guided by small RNA. Genome Res 23:1651–1662

Rivin CJ, Grudt T (1991) Abscisic acid and the developmental regulation ofembryo storage proteins in maize. Plant Physiol 95: 358–365

Roberts A, Pimentel H, Trapnell C, Pachter L (2011) Identification of noveltranscripts in annotated genomes using RNA-Seq. Bioinformatics 27:2325–2329

Shepard PJ, Hertel KJ (2008) Conserved RNA secondary structures pro-mote alternative splicing. RNA 14: 1463–1469

Staiger D, Brown JW (2013) Alternative splicing at the intersection of biologicaltiming, development, and stress responses. Plant Cell 25: 3640–3656

Stamm S, Ben-Ari S, Rafalska I, Tang Y, Zhang Z, Toiber D, Thanaraj TA,Soreq H (2005) Function of alternative splicing. Gene 344: 1–20

Thatcher SR, Zhou W, Leonard A, Wang BB, Beatty M, Zastrow-Hayes G,Zhao X, Baumgarten A, Li B (2014) Genome-wide analysis of alternativesplicing in Zea mays: landscape and genetic regulation. Plant Cell 26:3472–3487

Trapnell C, Williams BA, Pertea G, Mortazavi A, Kwan G, van Baren MJ,Salzberg SL, Wold BJ, Pachter L (2010) Transcript assembly andquantification by RNA-Seq reveals unannotated transcripts and isoformswitching during cell differentiation. Nat Biotechnol 28: 511–515

Wang W, Vinocur B, Shoseyov O, Altman A (2004) Role of plant heat-shock proteins and molecular chaperones in the abiotic stress response.Trends Plant Sci 9: 244–252

Wu Y, Deng Z, Lai J, Zhang Y, Yang C, Yin B, Zhao Q, Zhang L, Li Y, YangC, et al (2009) Dual function of Arabidopsis ATAF1 in abiotic and bioticstress responses. Cell Res 19: 1279–1290

Xie DX, Feys BF, James S, Nieto-Rostro M, Turner JG (1998) COI1: anArabidopsis gene required for jasmonate-regulated defense and fertility.Science 280: 1091–1094

Xu L, Liu F, Lechner E, Genschik P, Crosby WL, Ma H, Peng W, Huang D,Xie D (2002) The SCF(COI1) ubiquitin-ligase complexes are required forjasmonate response in Arabidopsis. Plant Cell 14: 1919–1935

Yan K, Liu P, Wu CA, Yang GD, Xu R, Guo QH, Huang JG, Zheng CC(2012) Stress-induced alternative splicing provides a mechanism for theregulation of microRNA processing in Arabidopsis thaliana. Mol Cell48: 521–531





Documents

Genome-Wide Analysis of Alternative Splicing during ... · Genome-Wide Analysis of Alternative Splicing during Development and Drought Stress in Maize1[OPEN] Shawn R. Thatcher, Olga